Recognition of Daily Activity in Living Space based on Indoor Ambient Atmosphere and Acquiring Localized Information for Improvement of Recognition Accuracy

Abstract

The system watching over elder's life is very important in a super-aged society Japan. In this paper, we describe a method to recognize resident's daily activities by means of using the information of indoor ambient atmosphere changes. The measuring targets of environmental changes are of gas and smell, temperature, humidity, and brightness. Those changes have much relation with resident's daily activities. The measurement system with 7 sensors (4 gas sensors, a thermistor, humidity sensor, and CdS light sensor) was developed for getting indoor ambient atmosphere changes. Some measurements were done in a one-room type residential space. 21 dimensional activity vectors were composed for each daily activity from acquired data. Those vectors were classified into 9 categories that were main activities by using Self-Organizing Map (SOM) method. From the result, it was found that the recognition of main daily activities based on information on indoor ambient atmosphere changes is possible. Moreover, we also describe the method for getting information of local gas and smell environmental changes. Gas and smell environmental changes are related with daily activities, especially very important action, eating and drinking. And, local information enables the relation of the place and the activity. For such a purpose, a gas sensing module with the operation function that synchronizes with human detection signal was developed and evaluated. From the result, the sensor module had the ability to acquire and to emphasize local gas environment changes caused by the person's activity.